Conflict of Interest Disclosures
Once a year your board members fill in a form, most of them write "none", and someone puts the stack in a drawer until the auditor asks. The form is not the control. The control is whether anyone ever compares those disclosures against who your organization actually pays, and almost nobody does, because it is tedious cross-referencing work that requires knowing things about your directors that nobody has written down.
The IRS asks on the Form 990 whether your organization has a written conflict of interest policy, whether officers, directors, and key employees are required to disclose annually, and whether the organization regularly and consistently monitors and enforces compliance with it. Most nonprofits answer yes to all three. The first two are usually true. The third is frequently aspirational, and the gap between the second and third is where the actual risk sits.
Nonprofit conflicts of interest are rarely dramatic. The overwhelming majority are ordinary and often benign: a board member whose firm provides a service at a fair price, a director who also sits on the board of a grantee, a treasurer whose spouse works at the bank holding your accounts. Most are fine. The problem is almost never the conflict, it is the failure to surface it, evaluate it, and record what the board decided.
The consequences of getting it wrong scale with the transaction. At the mild end, an auditor asks questions and a Schedule L gets filed late. At the serious end, the intermediate sanctions rules under Internal Revenue Code section 4958 impose excise taxes on individuals who receive excess benefits from a nonprofit, and on the organization managers who knowingly approved them. Personal liability for a board member is an unpleasant surprise for a volunteer who thought they were helping.
This is a good candidate for AI assistance because the underlying work is comparison across sources that nobody has time to reconcile by hand. What follows covers what a disclosure process needs to capture, why annual forms miss most real conflicts, exactly where AI helps with detection, the hard limits on what it should ever conclude, and how to run the board conversation when something surfaces.
What the Annual Form Should Actually Ask
Most disclosure forms ask a version of "do you have any conflicts of interest?" and then rely on the signer to know what counts. This is asking a volunteer to make a legal determination, and it reliably produces the answer "no" from people who have a disclosable relationship they did not recognize as one. A better form asks about facts rather than conclusions.
Ask what organizations the person works for, owns, or has a material financial interest in. Ask what other boards they serve on. Ask about their immediate family members' employment and business interests. Ask whether they or a family member has any business relationship with your organization, has received anything of value from it, or is employed by it. Ask about relationships with your vendors, your grantees, your funders, and your peer organizations. Then let the governance committee determine what constitutes a conflict, which is their job rather than the discloser's.
The scope of who completes the form matters as well. Directors and officers are the obvious population, and key employees as defined for Form 990 purposes should be included. Many organizations extend it to anyone with purchasing authority, anyone involved in vendor selection, and program staff who determine who receives services or subawards. The people who can direct organizational resources are the population that matters, and that is often broader than the board.
Finally, the form is the floor rather than the ceiling. A workable policy requires disclosure when a conflict arises, not only at annual signing, and requires it before the board acts on the matter rather than afterward. Recusal from both the discussion and the vote should be the norm for anything material, and the minutes need to reflect it, which is one of the reasons how minutes are produced matters more than it appears.
Ask for facts, not conclusions
What a useful disclosure form collects
- Employer, ownership interests, and material financial interests
- Other board and advisory positions held
- Immediate family employment and business interests
- Any transaction with the organization in the past year
- Relationships with vendors, grantees, funders, and peer organizations
- Gifts, discounts, or benefits received in connection with the role
Who should complete one
Anyone who can direct resources
- All directors and officers, without exception
- Key employees as defined for Form 990 reporting
- Anyone with purchasing or contracting authority
- Staff who participate in vendor selection or grantmaking
- Committee members who are not directors, including advisory members
Why the Annual Form Misses Most Real Conflicts
Understanding the failure modes explains why detection has to come from somewhere other than the form itself.
The first is that people do not recognize their own conflicts. A board member whose brother-in-law owns the landscaping company your facility uses does not think of that as their conflict, because it is not their business and they had nothing to do with the selection. They answer no in complete good faith. Disclosure depends on self-identification, and self-identification is unreliable precisely where relationships are indirect.
The second is timing. Forms are signed in January and the vendor decision happens in August, by which point the signed form is eight months stale and nobody is looking at it anyway. Board composition changes, people change jobs, family circumstances shift, and the disclosure captured a moment that has passed.
The third is that nobody reads them. The forms are collected, confirmed as complete, and filed. Whether anyone compares the disclosures against the vendor list, the grantee list, or the payroll is a separate question, and in most organizations the answer is no. This is the step that converts disclosure into monitoring, and it is the one that gets skipped.
The fourth is scope drift. A conflict disclosed and appropriately managed in year one is often forgotten by year three, particularly if the person who managed it has left. The relationship persists, the transaction volume may have grown considerably, and the original approval no longer covers what is actually happening.
The fifth is social. Small nonprofit boards are often built from overlapping networks, and raising a question about a respected colleague's relationship feels like an accusation. Where the culture treats disclosure as suspicion rather than as routine hygiene, people disclose less and the process quietly hollows out.
Where AI Helps: Comparison, Not Judgment
Every failure above is a comparison problem. The information needed to spot most conflicts already exists inside your organization, distributed across the disclosure forms, the vendor master, the payroll, the grantee list, and the board roster. Nobody has ever put those side by side, because doing it by hand for thirty people against six hundred vendors is a week nobody has.
Disclosures against the vendor list. The core check. Comparing employers, family employers, and business interests named on disclosure forms against every entity your organization has paid in the past year catches the relationships people did not think to mention. Name variations, subsidiaries, and doing-business-as names make this genuinely hard manually and straightforward with the right tooling.
Disclosures against grantees and program partners. Equally important for grantmaking organizations and fiscal sponsors, and frequently overlooked. A director who serves on the board of an organization applying for your funds has a conflict regardless of how modest the grant is.
Shared surnames and addresses across payroll and vendors. A blunt instrument that catches real things. Family employment relationships and vendors registered to a staff member's home address are exactly the pattern that appears in the sector's embezzlement cases, and both are visible in data you already hold.
Year-over-year change detection. Comparing this year's disclosures against last year's surfaces new employment, new board seats, and dropped disclosures. A relationship that appeared for three years and vanished in the fourth is worth a question, and nobody notices that by reading a single year's stack.
Schedule L preparation. Form 990 Schedule L requires reporting of certain transactions with interested persons, including loans, grants or assistance, and business transactions. Assembling that from disclosures plus transaction data is exactly the kind of cross-source compilation that gets rushed at filing time. Doing it as a standing process rather than an annual scramble also means you find out about a reportable transaction while you can still manage it. Our guide to drafting Form 990 narrative sections covers the adjacent work.
Meeting agenda screening. The most useful forward-looking application. Before each board meeting, checking agenda items against the disclosure register to identify who may need to recuse turns recusal from something remembered in the moment into something prepared for. It also removes the social awkwardness, since the recusal is procedural rather than raised by a colleague.
Cross-checks worth running annually
Each produces questions, never conclusions
- Disclosed employers and business interests against the full vendor list
- Disclosed board seats against your grantee and applicant lists
- Surnames and addresses shared between payroll and vendor records
- This year's disclosures against last year's, flagging additions and removals
- Vendors added in the past year without a documented selection process
- Previously disclosed relationships where transaction volume has grown materially
- Anyone who owes a disclosure form and has not submitted one
The Limits, Which Matter More Here Than Almost Anywhere
Everything above produces questions. None of it produces answers, and the distinction is not academic. A flagged item is an allegation about a volunteer's integrity if it is treated as a finding, and getting that wrong damages people and relationships in ways that are hard to repair.
The name-matching problem alone demands care. Common surnames produce coincidental matches constantly. Two people named Rodriguez in a database of four hundred vendors and thirty staff is not evidence of anything. Any match is a starting point for a private conversation, never an entry in a report circulated to the board.
A second limit is that whether a relationship constitutes a conflict, and what to do about it, is a judgment reserved to the governance committee or the board. Materiality, the nature of the interest, whether the transaction was at fair value, and whether the process was independent are all matters requiring people who know the context and who carry the fiduciary responsibility. A tool that scores conflicts and recommends outcomes has assumed a role that is not available to it.
Third, resist the temptation to widen the net with external data. It is now technically easy to search public records, corporate registrations, property records, and social connections for every board member. Doing so without their knowledge is investigating volunteers, and it will destroy trust the moment it becomes known. Conflict of interest processes work on disclosure and good faith. Where you have specific cause for concern, that is a matter for counsel and a deliberate decision by the board, not a routine automated sweep.
Fourth, treat the data itself carefully. Disclosure forms contain personal financial and family information about volunteers who provided it for a narrow purpose. Access should be limited to the governance committee and whoever administers the process, retention should be defined, and the material should not be routed through tooling without confirming what happens to it. Our guide to privacy risk assessment for nonprofit AI projects covers how to evaluate that, and the reasoning in our discussion of board-level AI risk registers applies directly.
Fifth, be transparent with the people disclosing. Board members should be told that disclosures are compared against vendor and grantee records, because that is a legitimate governance practice and saying so plainly increases the quality of disclosure. Discovering after the fact that their form was run through an automated screen they were not told about is the kind of surprise that costs you a director.
Appropriate
Comparison of records you already hold
- Cross-checking internal records against submitted disclosures
- Flagging items for a private conversation with the individual
- Screening board agendas against the disclosure register
- Assembling Schedule L candidates for review
- Tracking outstanding forms and expiring approvals
Not appropriate
Judgment and investigation
- Concluding that a conflict exists, or scoring its severity
- Recommending recusal, removal, or termination of a relationship
- Searching external and public records about volunteers routinely
- Circulating unverified name matches to the full board
- Running any of it without telling the people who disclosed
What Happens When Something Surfaces
Detection is the easy half. The half that determines whether any of this works is what the organization does when a flag turns out to be real, and most nonprofits handle that badly because they have never decided in advance how it goes.
Start privately. The board chair or governance committee chair speaks with the individual before anything goes further, and the framing is a question rather than an accusation: our annual review noticed a possible connection, can you help us understand it. Most of the time the explanation resolves it in a sentence, and handling it this way costs nothing when the flag was a coincidence.
Where a real conflict exists, evaluate it rather than reflexively eliminating it. Many conflicts are manageable and some are unavoidable, particularly in small communities and specialized fields where the people who understand your work are the same people who work in it. The questions are whether the transaction is at fair value, whether the process for selecting it was independent of the conflicted person, whether the arrangement is documented, and whether it can be disclosed comfortably.
Then follow the process precisely and record it. The conflicted person discloses, answers questions, and leaves the room. The remaining directors discuss and vote. The minutes record the disclosure, the departure and return, the abstention, and the basis for the decision, including any comparability data considered. That record is the entire value of the process. A well-managed conflict with no documentation looks identical to an unmanaged one when an auditor arrives three years later.
Set a review date rather than approving indefinitely. A vendor relationship approved in 2023 at modest volume should be revisited, and the approval should carry an expiry. Standing approvals are how manageable conflicts become uncomfortable ones without anyone deciding.
And build the culture that makes all of this work. Boards where the chair discloses their own relationships first, where disclosure is treated as ordinary rather than as an admission, and where recusal is unremarkable get far better information than boards where the topic is tense. This is a governance culture question as much as a process one, and it connects to the broader thinking in our guide to board oversight in an AI-enabled organization.
What the record should show for any managed conflict
This is what protects the organization and the individual
- The nature of the interest, disclosed before the board acted
- That the conflicted person left the discussion and the vote
- What alternatives were considered, if any
- The basis for concluding the terms were fair and reasonable
- The decision, the vote, and any abstentions
- A date when the arrangement will be reviewed again
Conclusion
Conflict of interest compliance in most nonprofits stops at collection. The forms come in, they are filed, and the box on the 990 gets ticked. The monitoring the IRS asks about, and that the policy actually promises, requires comparing what people disclosed against what the organization actually did, and that comparison has never been affordable at the scale most organizations need.
It is affordable now. Cross-referencing disclosures against vendors, grantees, payroll, and prior years is exactly the kind of tedious multi-source comparison that AI does well and that no volunteer committee was ever going to do by hand. That single addition turns a filing exercise into a functioning control, and it catches the indirect relationships that self-disclosure structurally misses.
Keep the boundary clear. The tool finds candidates for a conversation. It does not decide whether a conflict exists, how serious it is, or what to do, and it does not go looking through public records about your volunteers. Tell your board what you are doing, handle the first conversation privately and generously, document what you decide, and set a date to look again. Done that way, this is one of the cheapest governance improvements available to a nonprofit board.
Turn Your Disclosure Forms Into an Actual Control
We help nonprofit boards use AI on the governance work that has always been too tedious to do properly, with the boundaries that keep trust intact.